On Friday, David Sacks, the former White House AI and crypto czar and current co-chair of the President's Council of Advisors on Science and Technology, waded into the open-source AI debate with a clear message: don't kneecap the open models.

His proof point? Harvey, a legal-tech startup that just unveiled its new model, Tenet. Sacks sees it as a prime example of American ingenuity built on open systems.

Sacks Warns Against Restrictions

In a post on X, Sacks praised Harvey's approach, saying it shows how "American companies are building world-class specialized models."

Here's the interesting part: Harvey took an open-source base model, Kimi K3, and trained it on legal data. The result? Top-tier performance on legal benchmarks at a fraction of the cost of proprietary "frontier" models.

Sacks didn't mince words about the implications. "Restrictions that kneecap open models would do nothing to stop Chinese labs from shipping the next Kimi," he argued. Instead, such limits would cripple startups like Harvey that rely on open systems to build affordable, high-performance tools.

He also hinted that some closed AI labs might actually welcome these restrictions, since they'd eliminate competition.

This is a notable stance for Sacks, who earlier this year, along with investor Bill Ackman, called Kimi K3's benchmark rise "concerning" in July. That was when the model from Chinese startup Moonshot briefly outranked Anthropic's Claude Fable 5. Kimi K3 is a massive 2.8 trillion-parameter open-weight model that Moonshot launched in July.

Harvey's New Model Targets Legal Work Efficiency

Harvey unveiled Tenet this week, built on Kimi K3 and post-trained with Fireworks AI. According to Harvey, the model posted an 82% improvement in all-pass rate on the LAB legal benchmark and achieved state-of-the-art results on LAB Contracts.

Even better, Tenet runs at under one-fourth the cost of leading foundation models. Harvey plans to expand the system to more legal jurisdictions and practice areas.

The debate around open-source AI models has been heating up. Last week, Meta Platforms, Inc. (META) CEO Mark Zuckerberg continued advocating for open AI models, arguing that wider access can accelerate innovation and prevent a small number of companies from controlling the AI ecosystem.

Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by MarketDash editors.

Image via Shutterstock